Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “compute in memory”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 487 records · Page 27
All sourcesNASA NTRSDOE OSTI
DOE OSTI2025

The Artificial Scientist: in-Transit Machine Learning of Plasma Simulations

Large-scale simulations or scientific experiments produce petabytes of data per run. This poses massive challenges for I/O and storage when scientific analysis workflows are run manually offline. Unsupervised deep learning-based techniques to extract patterns and non-linear relations from these large amounts of data provide a way to build scientific understanding from raw data, reducing the need for manual pre-selection of analysis steps, but require exascale compute and memory to process the full dataset available. In this paper, we demonstrate a heterogeneous streaming workflow in which plasma simulation data is streamed directly to a Machine Learning (ML) application training a model on the simulation data in-transit, completely circumventing the capacity-constrained filesystem bottleneck. This workflow employs openPMD to provide a high level interface to describe scientific data and also uses ADIOS2, to transfer volumes of data that exceed the capabilities of the filesystem. We employ experience replay to avoid catastrophic forgetting in learning from this non-steady state process in a continual manner and adapt it to improve model convergence while learning in-transit. As a proof-of-concept, we approach the ill-posed inverse problem of predicting particle dynamics from radiation in a particle-incell (PIConGPU) simulation of the Kelvin-Helmholtz instability (KHI). We detail hardware-software co-design challenges as we scale PIConGPU to full Frontier, the Top-1 system as of June 2024 Top500 list.

Kelling, Jeffrey [Helmholtz-Zentrum Dresden Rossen↗
DOE OSTI2024

Bridging the Gap Between LLMs and LNS with Dynamic Data Format and Architecture Codesign

Deep neural networks (DNNs) have achieved tremendous success in the past few years. However, their training and inference demand exceptional computational and memory resources. Quantization has been shown as an effective approach to mitigate the cost, with the mainstream data types reduced from FP32 to FP16/BF16 and recently FP8 in the latest NVIDIA H100 GPUs. With increasingly aggressive quantization, however, the conventional floating-point formats suffer from limited precision in representing numbers around zero. Recently, NVIDIA demonstrated the potential of using a Logarithmic Number System (LNS) for the next generation of tensor cores. While LNS mitigates the hurdles in representing small numbers, in this work we observed a mismatch between LNS and the emerging Large Language Models (LLM), where LLM exhibits significant outliers when directly adopting the LNS format. In this paper, we present a data-format/architecture codesign to bright this gap. On the format side, we propose a dynamic LNS format to flexibly represent outliers at a higher precision, by exploiting asymmetry in the LNS representation and identifying outliers through a per-vector basis. On the architecture side, for demonstration, we realize the dynamic LNS format in a systolic array, which can handle the irregularity of the outliers at runtime. We implement our approach on an Alveo U280 FPGA as a prototype. Experimental results show that our design can effectively handle the outliers and resolve the mismatch between LNS and LLM, contributing to an accuracy improvement of 15.4% and 16% over the floating-point and the original LNS baselines, using four state-of-the-art LLM models. Our observation and design lay a solid foundation for the large-scale adoption of the LNS format in the next-generation deep learning hardware.

Haghi, Pouya↗
DOE OSTI2024

BCSR on GPU: A Way Forward Extreme-scale Graph Processing on Accelerator-enabled Frontier Supercomputer

Handling large graphs in a distributed environment requires effective partitioning across processors and efficient management of local partitions. In 2D partitioning, local graphs often become too sparse, making memory-efficient data structures crucial. Using the Compressed Sparse Row (CSR) format wastes space, especially for > 83% of vertices with empty edges for the sparse graphs. This study explores bit-CSR (BCSR), a modified CSR representation, on GPUs to reduce memory usage in graph computations. We achieved 16.67% memory savings on a sparse rmat dataset with 268 million vertices and 357 million edges, without performance degradation, supported by both theoretical and experimental storage savings of 33%. However, we observed a 1.7× slowdown in degree lookup times due to bitwise operations on AMD CPUs. This analysis highlights the potential of BCSR on GPUs for improving Graph500 benchmark performance on GPU-accelerated systems, such as the Frontier supercomputer.

Sattar, Naw Safrin↗
DOE OSTI2021

A Butterfly-Accelerated Volume Integral Equation Solver for Broad Permittivity and Large-Scale Electromagnetic Analysis

In this work, a butterfly-accelerated volume integral equation (VIE) solver is proposed for fast and accurate electromagnetic (EM) analysis of scattering from heterogeneous objects. The proposed solver leverages the hierarchical off-diagonal butterfly (HOD-BF) scheme to construct the system matrix and obtain its approximate inverse, used as a preconditioner. Complexity analysis and numerical experiments validate the O(N log 2 N) construction cost of the HOD-BF-compressed system matrix and O(N log 1.5 N) inversion cost for the preconditioner, where N is the number of unknowns in the high-frequency EM scattering problem. For many practical scenarios, the proposed VIE solver requires less memory and computational time to construct the system matrix and obtain its approximate inverse compared to a H matrix-accelerated VIE solver. The accuracy and efficiency of the proposed solver have been demonstrated via its application to the EM analysis of large-scale canonical and real-world structures comprising of broad permittivity values and involving millions of unknowns.

42 ENGINEERING↗
DOE OSTI2021

Penetration Through Slots in Overmoded Cavities

A resonant cavity undergoes three distinct behaviors with increasing frequency: 1) fundamental modes, localized in frequency with well defined modal distribution; 2) undermoded region, where modes are still separated, but are sufficiently perturbed by small imperfections that their spectral positions (and distributions) are statistical in nature; and 3) overmoded region, where modes overlap, field distributions follow stochastic distributions, and the slot acts as if in free space. Understanding the penetration through slots in the overmoded region is of great interest, and is the focus of this article. Since full-wave solvers may not be able to provide a timely answer for very high frequencies due to a lack of memory and/or computation resources, we develop bounding methods to estimate worst-case average and maximum fields within the cavity. Finally, after discussing the bounding formulation, we compare its results to full-wave simulations at the first, second, and third resonance supported by the slot in the case of a cylindrical cavity. Note that the bounding formulation indicates that results are nearly independent of cavity shape: only the cavity volume, frequency, and cavity quality factor affect the overmoded region, making this formulation a powerful tool to assess electromagnetic interference and electromagnetic compatibility effects within cavities.

overmoded cavity↗
DOE OSTI2023

Sparse Symmetric Format for Tucker Decomposition

Tensor-based methods are receiving renewed attention in recent years due to their prevalence in diverse real-world applications. There is considerable literature on tensor representations and algorithms for tensor decompositions, both for dense and sparse tensors. Many applications in hypergraph analytics, machine learning, psychometry, and signal processing result in tensors that are both sparse and symmetric, making them an important class for further study. Similar to the critical Tensor Times Matrix chain operation (TTM c ) in general sparse tensors, the $\underline{S}$ parse $\underline{S}$ ymmetric $\underline{T}$ ensor $\underline{T}$ imes $\underline{S}$ ame $\underline{M}$ atrix $\underline{c}$ hain (S 3 TTM c ) operation is compute and memory intensive due to high tensor order and the associated factorial explosion in the number of non-zeros. We present the novel Compressed Sparse Symmetric (CSS) format for sparse symmetric tensors, along with an efficient parallel algorithm for the S 3 TTM c operation. We theoretically establish that S 3 TTM c on CSS achieves a better memory versus run-time trade-off compared to state-of-the-art implementations, and visualize the variation of the performance gap over the parameter space. We demonstrate experimental findings that confirm these results and achieve up to 2.72× speedup on synthetic and real datasets. The scaling of the algorithm on different test architectures is also showcased to highlight the effect of machine characteristics on algorithm performance.

42 ENGINEERING↗
DOE OSTI2024

Field-free deterministic switching of all–van der Waals spin-orbit torque system above room temperature

Two-dimensional van der Waals (vdW) magnetic materials hold promise for the development of high-density, energy-efficient spintronic devices for memory and computation. Recent breakthroughs in material discoveries and spin-orbit torque control of vdW ferromagnets have opened a path for integration of vdW magnets in commercial spintronic devices. However, a solution for field-free electric control of perpendicular magnetic anisotropy (PMA) vdW magnets at room temperatures, essential for building compact and thermally stable spintronic devices, is still missing. Here, we report a solution for the field-free, deterministic, and nonvolatile switching of a PMA vdW ferromagnet, Fe 3 GaTe 2 , above room temperature (up to 320 K). We use the unconventional out-of-plane anti-damping torque from an adjacent WTe 2 layer to enable such switching with a low current density of 2.23 × 10 6 A cm –2 . This study exemplifies the efficacy of low-symmetry vdW materials for spin-orbit torque control of vdW ferromagnets and provides an all-vdW solution for the next generation of scalable and energy-efficient spintronic devices.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗
DOE OSTI2021

Efficient Parallel Sparse Symmetric Tucker Decomposition for High-Order Tensors

Tensor based methods are receiving renewed attention in recent years due to their prevalence in diverse real-world applications. There is considerable literature on tensor representations and algorithms for tensor decompositions, both for dense and sparse tensors. Many applications in hypergraph analytics, machine learning, psychometry, and signal processing result in tensors that are both sparse and symmetric, making it an important class for further study. Similar to the critical Tensor Times Matrix chain operation (TTMc) in general sparse tensors, the Sparse Symmetric Tensor Times Same Matrix chain (S3TTMc) operation is compute and memory intensive due to high tensor order and the associated factorial explosion in the number of non-zeros. In this work, we present a novel compressed storage format CSS for sparse symmetric tensors, along with an efficient parallel algorithm for the S3TTMc operation. We theoretically establish that S3TTMc on CSS achieves a better memory versus run-time trade-off compared to state-of-the-art implementations. We demonstrate experimental findings that confirm these results and achieve up to 2.9× speedup on synthetic and real datasets.

Shivakumar, Shruti↗
DOE OSTI2022

Towards Compact Neural Networks via End-to-End Training: A Bayesian Tensor Approach with Automatic Rank Determination

Post-training model compression can reduce the inference costs of deep neural networks, but uncompressed training still consumes enormous hardware resources and energy. To enable low-energy training on edge devices, it is highly desirable to directly train a compact neural network from scratch with a low memory cost. Low-rank tensor decomposition is an effective approach to reduce the memory and computing costs of large neural networks. However, directly training low-rank tensorized neural networks is a very challenging task because it is hard to determine a proper tensor rank a priori, and the tensor rank controls both model complexity and accuracy. Here, this paper presents a novel end-to-end framework for low-rank tensorized training. We first develop a Bayesian model that supports various low-rank tensor formats (e.g., CANDECOMP/PARAFAC, Tucker, tensor-train, and tensor-train matrix) and reduces neural network parameters with automatic rank determination during training. Then we develop a customized Bayesian solver to train large-scale tensorized neural networks. Our training methods shows orders-of-magnitude parameter reduction and little accuracy loss (or even better accuracy) in the experiments. On a very large deep learning recommendation system with over 4.2 ×10 9 model parameters, our method can reduce the parameter number to 1.6 ×10 5 automatically in the training process (i.e., by 2.6 ×10 4 times) while achieving almost the same accuracy. Code is available at https://github.com/colehawkins/bayesian-tensor-rank-determination.

compact neural networks↗
DOE OSTI2021

PAGANI: a parallel adaptive GPU algorithm for numerical integration

We present a new adaptive parallel algorithm for the challenging problem of multi-dimensional numerical integration on massively parallel architectures. Adaptive algorithms have demonstrated the best performance, but efficient many-core utilization is difficult to achieve because the adaptive work-load can vary greatly across the integration space and is impossible to predict a priori. Existing parallel algorithms utilize sequential computations on independent processors, which results in bottlenecks due to the need for data redistribution and processor synchronization. Our algorithm employs a high-throughput approach in which all existing sub-regions are processed and sub-divided in parallel. Repeated sub-region classification and filtering improves upon a brute-force approach and allows the algorithm to make efficient use of computation and memory resources. A CUDA implementation shows orders of magnitude speedup over the fastest open-source CPU method and extends the achievable accuracy for difficult integrands. Our algorithm typically outperforms other existing deterministic parallel methods.

Sakiotis, Ioannis↗
DOE OSTI2025

RingX: Scalable Parallel Attention for Long-Context Learning on HPC

The attention mechanism has become foundational for remarkable AI breakthroughs since the introduction of the Transformer, driving the demand for increasingly longer context to power frontier models such as large-scale reasoning language models and high-resolution image/video generators. However, its quadratic computational and memory complexities present substantial challenges. Current state-of-the-art parallel attention methods, such as ring attention, are widely adopted for long-context training but utilize a point-to-point communication strategy that fails to fully exploit the capabilities of modern HPC network architectures. In this work, we propose ringX, a scalable family of parallel attention methods optimized explicitly for HPC systems. By enhancing workload partitioning, refining communication patterns, and improving load balancing, ringX achieves up to 3.4 × speedup compared to conventional ring attention on the Frontier supercomputer. Optimized for both bi-directional and causal attention mechanisms, ringX demonstrates its effectiveness through training benchmarks of a Vision Transformer (ViT) on a climate dataset and a Generative Pre-Trained Transformer (GPT) model, Llama3 8B. Our method attains an end-to-end training speedup of approximately 1.5 × in both scenarios. To our knowledge, the achieved 38% model FLOPs utilization (MFU) for training Llama3 8B with a 1M-token sequence length on 4,096 GPUs represents one of the highest training efficiencies reported for long-context learning on HPC systems. Our code implementation is available at https://github.com/jqyin/ringX-attention.

Yin, Junqi [ORNL] (ORCID:0000000338435520)↗
DOE OSTI2025

Minimizing CGYRO HPC Communication Costs in Ensembles with XGYRO by Sharing the Collisional Constant Tensor Structure

First-principles fusion plasma simulations are both compute and memory intensive, and CGYRO is no exception. The use of many HPC nodes to fit the problem in the available memory thus results in significant communication overhead, which is hard to avoid for any single simulation. That said, most fusion studies are composed of ensembles of simulations, so we developed a new tool, named XGYRO, that executes a whole ensemble of CGYRO simulations as a single HPC job. By treating the ensemble as a unit, XGYRO can alter the global buffer distribution logic and apply optimizations that are not feasible on any single simulation, but only on the ensemble as a whole. The main saving comes from the sharing of the collisional constant tensor structure, since its values are typically identical between parameter-sweep simulations. This data structure dominates the memory consumption of CGYRO simulations, so distributing it among the whole ensemble results in drastic memory savings for each simulation, which in turn results in overall lower communication overhead.

CGYRO↗
DOE OSTI2020

Reeber v2.0

Reeber is a library for shared- and distributed-memory parallel computation of merge trees

Morozov, Dmitriy↗
DOE OSTI2023

Performance portable ice-sheet modeling with MALI

High-resolution simulations of polar ice sheets play a crucial role in the ongoing effort to develop more accurate and reliable Earth system models for probabilistic sea-level projections. These simulations often require a massive amount of memory and computation from large supercomputing clusters to provide sufficient accuracy and resolution; therefore, it has become essential to ensure performance on these platforms. Many of today’s supercomputers contain a diverse set of computing architectures and require specific programming interfaces in order to obtain optimal efficiency. In an effort to avoid architecture-specific programming and maintain productivity across platforms, the ice-sheet modeling code known as MPAS-Albany Land Ice (MALI) uses high-level abstractions to integrate Trilinos libraries and the Kokkos programming model for performance portable code across a variety of different architectures. In this article, we analyze the performance portable features of MALI via a performance analysis on current CPU-based and GPU-based supercomputers. The analysis highlights not only the performance portable improvements made in finite element assembly and multigrid preconditioning within MALI with speedups between 1.26 and 1.82x across CPU and GPU architectures but also identifies the need to further improve performance in software coupling and preconditioning on GPUs. We perform a weak scalability study and show that simulations on GPU-based machines perform 1.24–1.92x faster when utilizing the GPUs. The best performance is found in finite element assembly, which achieved a speedup of up to 8.65x and a weak scaling efficiency of 82.6% with GPUs. We additionally describe an automated performance testing framework developed for this code base using a changepoint detection method. The framework is used to make actionable decisions about performance within MALI. We provide several concrete examples of scenarios in which the framework has identified performance regressions, improvements, and algorithm differences over the course of 2 years of development.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2023

MPACT 4.4 Theory Manual

MPACT is a three-dimensional (3D) full-core neutron transport code capable of calculating subpin power distributions. Calculations are based on the Boltzmann transport equation for neutron fluxes for problems in which the detailed geometrical configuration of fuel components such as the pellet and cladding are explicitly retained. The cross-section data needed for the neutron transport calculation are obtained directly from a multigroup cross section library, which has traditionally been used by lattice physics codes to generate few-group homogenized cross sections for nodal core simulators. Hence, MPACT assumes neither a priori homogenization nor group condensation for the full core spatial solution. The 3D MPACT transport solution can be obtained using the method of characteristics (MOC), which employs discrete ray tracing within each fuel pin. However, for practical reactor applications, the direct application of MOC to 3D core configurations requires an excessive amount of memory and computing time due to the very large number of rays. For practical 3D full-core calculations, MPACT commonly uses an approximate “2D/1D” method that treats the radial (x and y) variables differently from the axial (z) variable. In particular, the radial dependence of the solution is calculated using transport theory, and the axial dependence is calculated using diffusion or P 3 theory. The 2D/1D method requires the core to be divided into a vertical stack of axial slices with a thickness of Δ z ≈ 5–10 cm. Each axial slice is divided radially into coarse spatial cells with boundaries that usually constitute the pin cell boundaries, for which Δ x = Δ y ≈ 1.5 cm. Then, each coarse radial cell (pin cell) is divided into 50–100 fine radial cells, which resolve the angular flux in the fuel, cladding, and moderator regions.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗
DOE OSTI2024

Exploring the Adoption Challenges of Post-Quantum Cryptography in EV Charging Infrastructure

The rapid evolution of electric vehicle (EV) technology and the corresponding growth of the Electric Vehicle Charging Infrastructure (EVCI) brings to light significant cybersecurity concerns, notably in the context of emerging post-quantum computing capabilities. This report, prepared by Pacific Northwest National Laboratory (PNNL) under the U.S. Department of Energy contract, delves into the challenges associated with integrating Post-Quantum Cryptography (PQC) into EVCI to safeguard against potential quantum computing threats. Post-quantum computers will eventually be able to invalidate technologies secured through public key cryptography. As part of this effort, the primary gaps and challenges in the EVCI were investigated with a focus on comparing traditional algorithms against PQC algorithms. One of the notable findings was that the P-521 algorithm was frequently surpassed in performance by PQC algorithms. This document provides a thorough examination of the hurdles the industry can expect when transitioning to PQC within the EVCI, such as interoperability concerns, the computational and memory demands of PQC algorithms, and the organizational readiness for such a transition. It emphasizes the necessity of a forward-thinking approach to cybersecurity, advocating for early and strategic engagement among EVCI stakeholders to ensure a seamless and cost-effective migration to quantum-resistant cryptographic standards. Through this report, the authors aim to catalyze awareness and action among policymakers, industry leaders, and cybersecurity professionals towards fortifying the EVCI against emerging quantum threats, thereby securing the infrastructure essential for the future of electric mobility.

33 ADVANCED PROPULSION SYSTEMS↗